benefits of home currency invoicing -...
TRANSCRIPT
1
Benefits of Home Currency Invoicing§
Kazunobu HAYAKAWA
Bangkok Research Center, Institute of Developing Economies, Thailand
Nuttawut LAKSANAPANYAKUL
Science and Technology Development Program, Thailand Development Research Institute, Thailand
Taiyo YOSHIMI#
Department of Economics, Nanzan University, Japan
Abstract: We present a model of endogenous choice of import frequency and invoice currency to
infer the benefits of home currency invoicing. In the model, those benefits are represented by fixed
costs of exchange rate risk management paid by importers when they work with foreign currency
invoicing. Using a very detailed dataset on Thai imports, we show that those benefits of average
Thai importer range between 7.3% (1,500 USD) of one-time shipment value in our most
conservative specification and 17.1% (3,600 USD). Those benefits become larger when the share of
the export country currency in daily global foreign exchange market turnover is lower, or the export
country is one of the partners of Thailand’s regional trade agreements. Further, we find that
frequency of shipments is higher and the value per shipment is smaller for transactions priced in
buyers’ currency than those not priced in it. Our model provides a rationale for these empirical
findings.
JEL Classification: F10; F14; F31; F40
Keywords: Invoice currency; Frequency of trade; Shipment
1. Introduction
Running an international currency provides benefits for a country as it liberates
§ We have benefited from discussions with and comments of Shingo Iokibe, Hugh Patrick and David Weinstein. We would also like to thank the seminar participants at The Japan Society of International Economics Kansai, Research Institute of Economy and Industry (RIETI) and Center on Japanese Economy and Business (CJEB), Columbia Business School. All remaining errors are ours. This work was supported by the Japan Society for the Promotion of Science (JSPS), Grant-in-Aid for Young Scientists (A) (15H05393). # Corresponding author: Taiyo Yoshimi; Address: Department of Economics, Nanzan University, 18 Yamazato-cho, Showa-ku, Nagoya-shi, Aichi-ken 466-8673, Japan; Tel.: +81-52-832-3111, fax: +81-52-835-1444; Email: [email protected]
2
home agents from the risk of exchange-rate fluctuations. There has been a long debate
on this topic, and huge amount of estimations of those benefits has been provided. One
of the major aspects of currency internationalization is trade invoicing. If particular
trade is invoiced in the home currency, home agents do not need to pay attention to
exchange-rate risk. In contrast, when the international transaction is invoiced in foreign
currencies, traders generally utilize some ways of exchange-rate risk management, such
as forward exchange rates and currency options, by paying visible and invisible costs. In
short, currency internationalization contributes to reducing broadly-defined trade costs
for firms in the home country.
In this paper, we quantify the benefits of home currency invoicing (HCI) through
measuring the costs for exchange-rate risk management in the case of foreign currency
invoicing (FCI). Namely, the benefits of currency internationalization in terms of
reducing the trade costs are quantified. To do that, we shed light on the relationship
between trade frequency and uncertainty. As theoretically demonstrated in Bekes et al.
(2014), firms adjust to increased uncertainty by reducing their number of trade
shipments and increasing their size per shipment. In the case of FCI, firms’ future
revenue/payment for every shipment in terms of its own currency is exposed to
unexpected exchange-rate fluctuation. This specific type of uncertainty in FCI motivates
us to focus on the trade-off between HCI and FCI. Suppose that importers are invoicing
their international transactions in terms of a foreign currency. If payments come after
contracts and importers’ home currency is depreciating, future payments in importers’
home currency will become cheaper. This implies one of the advantages of FCI. On the
other hand, FCI requires firms to incur some costs for exchange-rate risk management
in every transaction. Using such trade-off in invoice currency, we identify the costs for
exchange-rate risk management in FCI.
Specifically, we construct a model of endogenous choice of import frequency and
invoice currency to infer those costs. We extend the model developed by Kropf and
Sauré (2014) by two ways. First, we focus on the optimization behavior of importers
rather than exporters as we like to examine how the adoption of buyers’ (importers’)
currency invoicing helps their just-in-time orders and how it is related to the benefits of
HCI. Second, we introduce endogenous choice of invoice currency. As a result, our
model proposes a way to infer the costs of FCI. Further, it also provides two new
propositions on the relationship of the invoice currency with import frequency and the
value per shipment. One is that import frequency of FCI importers is lower than that of
HCI ones under some conditions. This consequence becomes more likely when the
costs of exchange-rate risk management become larger. The other is that the value per
3
shipment of FCI importers increases with the costs of exchange-rate risk management.
In the empirical part, we infer the costs for such exchange-rate risk management
by employing the customs import data in Thailand during 2007 to 2011. Our dataset,
which is obtained from the Customs Office, the Kingdom of Thailand, is
transaction-level import data and covers all commodity imports in Thailand. It contains
not only the usual data items such as Harmonized System (HS) eight-digit code or
export country but also import firm identification code, customs clearing date, and
invoice currency. For example, we can see that some firms import the same commodity
from the same country under the same invoice currency multiple times within the same
date. The case of Thailand not only provides us the detailed data necessary for our
analysis but also shows the case of large costs for exchange-rate risk management since
its home currency is not internationalized. Therefore, there are a sufficient number of
import transactions under FCI. By applying these data to our theoretical model, we infer
fixed costs per import shipment in the cases of HCI and FCI. We also examine how such
costs are related to export country characteristics such as turnover of export country
currency. In addition, with this dataset, we empirically investigate the above two
propositions derived from our theoretical model.
Our paper is related to literatures on trade frequency and invoice currency. The
former literature is recently growing. Eaton et al. (2008) is an early micro-level study on
this literature and provides various basic statistics on the number and frequency of
export transactions in Columbia. Alessandria et al. (2010) demonstrate that the existence
of fixed costs per import shipment leads to the lumpiness of import transactions. The
calibration of their inventory model for Chilean import data shows that such fixed costs
amount “to approximately 3.6 percent of the average value of an import shipment”.
Kropf and Sauré (2014), which is the closest paper to ours, also compute fixed costs per
export shipment with Swiss export data and found a similar magnitude. In this paper, by
applying their method, we compute not only fixed costs per import shipment but also
those of exchange-rate risk management in the case of FCI. When importing under
home currency, firms do not need to bear the latter costs. Therefore, those costs are
interpreted as the benefits of HCI. To the best of our knowledge, such costs have never
been quantified by the way driven from micro-founded economic theory using highly
disaggregated dataset.
Furthermore, Hornok and Koren (2014) and Bekes et al. (2014) shed more light
on the correlation of shipment frequency with several variables. The former study found
in the export data in the U.S. and Spain that export shipments are less frequent and
larger when exporting to countries with larger per-shipment costs. Using French export
4
data, as mentioned above, the latter study shows that firms adjust to increased
uncertainty by reducing their number of shipments and increasing their shipment size.
Against these studies, our paper sheds light on the correlation of the shipment frequency
and size with a novel element, invoice currency. Nevertheless, as mentioned above, this
analysis will be similar to that in Bekes et al. (2014) because the choice of invoice
currency is related to one specific type of uncertainty, future unexpected exchange-rate
fluctuation. In this sense, our analysis can be compliment to Bekes et al. (2014).
The other related literature is one on the choice of invoice currency. Many
researchers have examined how traders determine the invoice currency and have tried to
endogenize its choice. For instance, Goldberg and Tille (2008) explore the major driving
forces for currency invoicing in international trade with a dataset covering 24 countries.
Further, Goldberg and Tille (2009) examine the choice of the invoice currency through
bargaining between importers and exporters. Based on a highly disaggregated dataset of
Canada, they show that larger transactions are more likely to be invoiced in the
importers’ currency, which is rationally explained by their model of endogenous choice
of invoice currency through a bargaining process. Gopinath et al. (2010) construct a
model of endogenous currency choice where importers choose the invoice currency so
that realized degree of exchange rate pass-through becomes close to its desired degree.
In this paper, on the other hand, we assume that forward exchange premiums are
exogenously offered by financial institutions to each firm and are heterogeneous across
firms. Under this setting, we will show that given the fixed cost of exchange rate risk
management, importers choose FCI when the home currency is significantly discounted
in the suggested forward exchange rate and thus the cost of imported intermediate
inputs is supposed to become large when the payment is settled. The advantage of our
framework is that it provides a way to infer the difference in broadly-defined trade costs
including exchange rate risk between HCI and FCI, which reveals practical benefits of
HCI in international transactions.
The level of disaggregation is also one of the advantages of this study. With our
data, we can derive detailed information including the frequency of transactions, value
per shipment, and the invoice currency for each transaction. Several recent studies also
utilize dataset with a comparable level of disaggregation with ours to examine topics on
invoice currencies. Chung (2014) reveals that exporters’ dependence on imported inputs
affects their choice of invoicing currency using a dataset which covers all United
Kingdom trade transactions with non-European Union (EU) countries. Fabling and
Sanderson (2015) examine how invoice currency is linked to the degree of exchange
rate pass-through with New Zealand’s data. Reiss (2015) uses Brazilian trade data, and
5
finds that the Brazil-Argentina policy of providing payment orders associated to an
exchange transaction between their currencies had significant impacts on the choice of
invoice currencies.
Also, our paper is related to a huge literature on the measurement of trade costs
because the costs of exchange-rate risk management for FCI are a part of the
broadly-defined trade costs. Anderson and van Wincoop (2004) is one of the most
influential papers in this literature. Rose and van Wincoop (2001) and the sequential
papers estimate the costs of not sharing the same currency in trading. To do that, these
papers mainly estimate the gravity equation. As a result, in Rose and van Wincoop
(2001), the tariff equivalent of those costs is estimated to 14 percent. Compared with
these studies, we compute the more specific costs when using different currency in
trading, i.e., fixed costs of exchange-rate risk management for FCI. In terms of
measuring the absolute magnitude of fixed costs, our paper is also related to Das et al.
(2007) and Cherkashin et al. (2015). These studies structurally estimated fixed costs for
exporting (around 400 thousand US dollars (USD)) and certifying the origin of exported
goods (around four thousand USD), respectively.
The major results of this paper are summarized as follows. Firstly, the benefits of
HCI for average Thai importer are positive and range between 7.3% (1,500 USD) of
one-time shipment value in our most conservative specification and 17.1% (3,600 USD).
This implies that the adoption of the home currency as an invoice currency provides the
benefits for home importers. Secondly, those benefits become larger when the share of
the export country currency in daily global foreign exchange market turnover is lower,
or the export country one of the partners of Thailand’s regional trade agreements (RTAs).
Thirdly, the frequency of import shipments is higher and the value per shipment is
smaller for transactions priced in importers’ currency than those not priced in it, which
is rationally interpreted by our theoretical model.
The rest of this paper is organized as follows. In Section 2, we present a model of
endogenous choice of import frequency and invoice currency. Section 3 provides a way
to infer the benefits of HCI based on our model. In Section 4, we examine the
propositions on the relations between invoice currency, import frequency, and the value
per shipment. In Section 5, we infer the fixed costs per shipment and the benefits of
HCI, and examine their determinants. Finally, Section 6 concludes the paper.
2. The Model This section presents a partial equilibrium model of endogenous choice of import
6
frequency and invoice currency. As we noted, the most closely-related study to ours is
Kropf and Sauré (2014). We extended their model by two ways. Firstly, we focus on the
optimization behavior of importers and, exporters are simply assumed to supply the
amount of intermediate goods upon the request of final-good producers. Secondly, the
choice of invoice currency is endogenous, which helps to figure out the simultaneous
determination of invoice currency and import frequency. Figure 1 present a simplified
image of our model.
=== Figure 1 ===
2.1. Representative Household Suppose the economy which is constructed with infinite number of countries.
Each country is indexed by and countries are distributed continuously in the range
0, 1 . In each country, the representative household, final- and intermediate-good
producers, and financial institutions exist. The representative household purchases final
goods from local producers and consume them. The final-good market is
monopolistically competitive, and each final-good producer supplies particular variety
which is indexed by . The number of final goods is infinite and is assumed to be
distributed in the range 0, 1 .
The representative household has a linear utility function over the consumption as
follows:
.
Here, is the consumption index which is defined by a Dixit-Stiglitz function over
monopolistic-competitive varieties in the following manner:
≡ ,1 ∞, 1
where is the consumption of the variety , and is the elasticity of substitution
between varieties. The optimal allocation of any given consumption of each variety of
goods yields the following demand functions:
, 2
where the consumer price index (CPI) is given by
≡ . 3
7
Here, we used the assumption ⁄ which implies that the representative
household is endowed with the exogenous amount of nominal income, , and expenses
it only to current consumption.
2.2. Forward Exchange Rates Each final-good producer makes its output using imported or domestically
produced intermediate inputs. Final-good producers determine the frequency of
transactions with intermediate-good producers. Those which import intermediate inputs
from abroad also determine the invoice currency. Particularly, we consider two
alternative cases of invoicing, HCI and FCI.1 The critical difference between these two
cases is that importers are free from the risk of exchange-rate fluctuations in the HCI
case but they have to manage it in the FCI case. Suppose that home importers and
foreign intermediate-good exporters make contracts of international transactions in the
current period, and the payment will be taken place after ′ periods. If the trade is
invoiced in importers’ (home) currency, they do not have to care about the risk of
exchange rate fluctuations over the gap between contract and payment. On the other
hand, they have to consider how they manage the risk when the trade is invoiced in
foreign currencies.
Let and ′ denote spot exchange rates of home currency in the current
period and after ′ periods, respectively, against the invoice currency which is used for
the trade between home importers and their intermediate-good suppliers. We assume
that each importer manages the risk of exchange rate fluctuations by utilizing the
forward exchange rate with the nominal fixed cost, , when the trade is not invoiced
in their home currency, i.e. in the HCI case. This cost includes the documentation cost
of forward exchange rate contracts, and the transaction cost charged by the financial
institutions. Thus, we interpret as the benefits of adopting the home currency as an
invoice currency in import. Further, we also assume the existence of the fixed costs per
shipment, . We define ′ as the forward exchange rate charged by the financial
institution which serves for the importer .
We assume the heterogeneity in the forward exchange premiums offered by
financial institutions to each company. It is natural to assume such heterogeneity
1 In studies such as Chung (2014) and Fabling and Sanderson (2015), vehicle currency invoicing and producer currency invoicing are identified. The purpose of this study is to reveal the benefits of HCI, which corresponds to local currency invoicing, relative to other two invoicing manners. Consequently, in the theoretical part, we jointly consider vehicle and producer currency invoicing as FCI for simplicity, and focus on the comparison between HCI and FCI. In the empirical part, we identify those two types of FCI.
8
because different financial institutions offer different premiums. Further, even the same
bank can offer different premiums depending on the relation with each company. For
instance, if a bank has a long relation with a particular company, it might provide
special discount for future currency exchanges. Or, if the owner of a company has a
knowledge or experience to find a better financial institution, it can make a better
forward exchange contract. We denote the forward premium offered by a financial
institution to each firm by , which is assumed to be constant over time. Thus, the
forward exchange rate is determined in the following manner:
′ . 4
2.3. Final-good Producers Each final-good producer has linear production function over imported
intermediate inputs and the productivity in the following manner:
. 5
Here, is the output of final-good producer , is the productivity, and is the
unit of imported or domestically produced intermediate inputs. In the model, we assume
that each final-good producer imports intermediate inputs from one counterpart for
simplicity, although the data shows that one company imports from multiple countries
in many cases. We can interpret our model so that multiple importers belong to one
company as an importing section and each of them independently determines how to
import each product. This interpretation leads to an economy where one company
imports from multiple countries, which is consistent with the data. We denote the
intermediate-good price denominated in the invoice currency by ∗ ≡ ̃∗. Here, is
the gross tariff rate, and 1 and 1 for final-good producers which use
imported and domestic intermediate inputs, respectively. ̃ ∗ is the mill price. Both
and ̃ ∗ are exogenous. Thus, the nominal marginal cost in the home currency is derived
as follows: ∗
. 6
Final-good producers make their output as soon as they import intermediate inputs. It
should be noted that the invoice currency is the home currency and 1 for
final-good producers which use domestic intermediate inputs.
When there is a time gap between final-good production and the sales, storage
costs take place. Following Kropf and Sauré (2014), we assume the existence of the
gross ad valorem storage costs, . Given the price markup, marginal cost, and
9
storage cost, the consumer price of the variety is derived as follows:2
′1
∗
. 7
The operating profit of per unit sales of each final good is derived in the following
manner: ∗
1∗ 1
1
∗
. 8
Combining (2) and (8), the operating profit of total sales at time is obtained as
1
1
∗
. 9
We denote the time between two shipments by ∆ . In other words, each
final-good producer sells their output over ∆ , and makes the order of next shipment
after all the inventory stock is sold out. Normalizing one year to unity, the interval ∆
is expressed as a fraction of years. Consequently, ∆ represents the import frequency
or the number of shipments per year between home country and the origin of imported
intermediate inputs. Taking and as given, the present value per shipment is
derived in the following manner:3
Π ∆ ≡∆ 1 ∆
1, 9
where,
≡1
1
∗
, 10
≡ . 11
Note that Δ ∈ 0,∞ → ∈ 0,1 .
2.4. Optimal Frequency of Shipments Import frequency and invoice currency are simultaneously determined. Firstly,
final-good producers calculate the optimal frequency of shipments in a given period in
HCI and FCI cases. Secondly, they compare present values of profits from all shipments
2 Sum of and can be interpreted as effective storage cost as future depreciation of home currency increases the cost of imported intermediate inputs denominated in the home currency. It is natural to assume 0, which implies storages cannot become beneficial in effect. This leads to a condition , where is the lower bound of . As a result, is supposed to follow a
probability distribution with the range [ , ∞). 3 In Appendix A, we will show that there is a distribution of shipment dates for which becomes constant.
10
in both cases, and choose the invoice currency which realizes better outcome. In this
subsection, we show the detail of these optimization processes. In the calculation of
optimal frequency of shipments, final-good producers face the trade-off between storage
costs and fixed costs per shipment. Given the total value of all shipments per period, the
higher frequency of shipments is associated with the larger value per shipment.
Consequently, total amount of fixed costs per shipment and storage costs in a given
period become larger (smaller) and smaller (larger), respectively, if importers make
more (less) frequent shipment orders.
We denote total fixed costs for each shipment charged to the importer by .
Recall that corresponds to and in the HCI and FCI cases, respectively.
Further, importers do not need to utilize forward exchange rates, and disappears
from equations in the case of HCI ( 0). The present value of profits from all
shipments is defined by
≡ Π ∆1
1
1
1 12
Each final-good producer calculates the optimal frequency by determining to
maximize in HCI and FCI cases.
The first order condition is derived as follows:
11
1 0 13
Given that the range of is ∈ 0,1 , the left hand side of (13) is proved to be
decreasing in . Further, it becomes positive and negative in the limit of → 0 and
→ 1, respectively, given the necessary condition for final-good producers’ market
entry that 0. Thus, is uniquely determined for both HCI ( ) and FCI ( )
cases (∆ and ∆ , henceforth). Note that the optimal frequency in the HCI case does
not depend on . Further, ( ) is positively related to the frequency 1 ∆⁄ (1 ∆⁄ ).
Two consequences are implied by the first order condition. First, the optimal
frequency is negatively associated with total fixed costs for each shipment, , for both
HCI and FCI cases. Thus, the existence of can be a source of lower frequency in the
FCI case than the HCI case. Second, in the case of FCI, the optimal frequency is
negatively associated with the forward premium, . Higher premium leads to the rise
11
of the cost of imported intermediate inputs denominated in the home currency. As a
result, for a final-good producer which faces higher premium, it is more difficult to
cover total costs of each shipment, and thus the frequency becomes lower.
2.5. Choice of the Invoice Currency Combining (12) and (13), the equilibrium value of is written by
. 14
Final-good producers compare the in HCI and FCI cases, and choose the invoice
currency which realizes better consequence. Thus, the condition for HCI invoicing is
, or
. 15
It is straightforward to show that the right hand side of this condition is decreasing in
. Thus, final-good producers with high value of forward premium tend to choose their
home currency as an invoice currency. The condition (15) determines the cutoff forward
premium for which it holds with equality. It is straightforward to show the
uniqueness of given the monotonic decrease of the right hand side of (15) in .
Figure 2 depicts the relation between the value of the forward premium and
choice of the invoice currency. In the vicinity of the cutoff forward premium, import
frequency of FCI importers is lower than that of HCI ones. The FCI curve shifts
downward when the fixed cost of utilizing forward exchange rate, , becomes larger.
This amplifies the likelihood of lower frequency for FCI importers than HCI ones. Thus,
we obtain the proposition on the relation between the invoice currency and frequency of
import shipments as follows:
Proposition 1: Import frequency of FCI importers is lower than that of HCI ones
around the cutoff forward premium, which is more likely when the fixed cost of
exchange-rate risk management is larger.
=== Figure 2 ===
Further, we examine the optimal value per shipment in terms of the home
currency. Combining (2) and (8), it is derived in the following manner:
12
≡ ∗∆ 1 1
1 . 16
We found that the optimal frequency becomes higher when the total fixed cost of each
shipment becomes lower. This implies, according to (16), that the existence of can
lead to the larger value per shipment for FCI importers than HCI ones. However, it
depends on parameter values whether the value per shipment for the former is larger
than the latter in sum, even in the vicinity of the cutoff forward premium, as the sign of
the partial derivative of with respect to depends on parameter values. In sum,
we derive the following proposition on the relation between the invoice currency and
the value per shipment:
Proposition 2: The existence of the fixed cost of exchange-rate risk management
increases the value per shipment of FCI importers over that of HCI importers.
2.6. Benefits of HCI By combining (13) and (16) to eliminate , we derive the indirect measure for
the total fixed cost of each shipment as follows:
17
≡1 1 1
18
≡ 1 11 1
19
Recall that is and , respectively, for HCI and FCI, and 0 for HCI.
Thu, the indirect measure of is derived by
. 20
Further, by eliminating , we infer the measure for in the following manner:
. 21
As we noted, is interpreted as the benefits of using the home currency as an invoice
currency in international transactions, more simply the benefits of home currency
invoicing.
3. Data Overview This section takes an overview of invoice currency, import frequency, and imports
per shipment in Thailand. As mentioned in the introductory section, our dataset is
13
transaction-level import data from 2007 to 2011 and covers all commodity imports in
Thailand. It contains customs clearing date, HS eight-digit code, export country, import
firm identification code, invoice currency, import values in Thai Baht (THB), import
quantity, and quantity unit.4 Since our dataset does not include information on export
firms, we differentiate import transactions according to import firm, HS eight-digit code,
export country, and quantity unit. For example, if a firm has two records on imports of
the same HS code from the same export country under different quantity units, we
regard these two records as showing the import from two different exporters. In addition,
in our dataset, there are some firms in which multiple import transactions are recorded
in the same HS eight-digit commodity, the same export country, the same invoice
currency, the same unit, and the same date. We aggregate these data at a daily basis and
thus regard this case as one transaction.
We start from the sample distributions of the number of shipments per year and
average imports per shipment in 2011, which are depicted in Figures 3 and 4,
respectively. The former figure shows that more than a half of all import transactions
have only one shipment per year. Compared with the case of developed countries shown
in Kropf and Sauré (2014) and Bekes et al. (2014), the share of transactions of one
shipment per year is very high. Also, as in the case of developed countries, the density
naturally decreases with the increase of the shipment number. Figure 4 shows that the
distribution of average imports per shipment seems to follow the log-normal
distribution.
=== Figures 3 and 4 ===
Table 1 reports the share of each type of invoice currency in total import, in terms
of import transactions and import values. The invoice currency includes local currency
(HCI) and non-local currency (FCI). While the former is Thai baht (THB) in our context,
the latter is further decomposed into producer currency (i.e., national currency in export
countries) and vehicle currency (i.e., neither home nor producer currencies). The table
shows that approximately 60% and 80% of import in Thailand are invoiced in the
vehicle currency in terms of the number of transactions and of values, respectively. This
fact is consistent with the well-known one that a significant part of international
transactions in Asia is invoiced in third vehicle currency, particularly, US dollar (USD).
4 We drop the transactions in which the information on invoice currency, export country, or quantity unit is missing. Therefore, the aggregated figures of our data may not be consistent with the official figures on imports.
14
The table also shows that the local currency (i.e., THB) invoicing is relatively minor to
vehicle and producer currency invoicing. Its share is less than 10 percentages in our
sample in terms of both the number of transactions and the values.
=== Table 1 ===
Shares of invoice currencies vary across sectors and regions. Table 2 presents the
share of each type of invoice currency by industry, in terms of the number of import
frequency and the import values. In almost all industries, as in the case of Table 1, the
local currency has the lowest share. In those industries, the share of vehicle currency is
higher than that of producer currency in terms of both the number of import frequency
and the import values. In particular, footwear and textile industries have significantly
high share of vehicle currency. In addition, the extremely high share of vehicle currency
in mineral products in terms of values reflects the fact that the prices of such mineral
products are determined in the international market in USD. On the other hand, the high
share of local currency can be found only in art products probably because of the strong
bargaining power of importers (buyers) in this industry.
=== Table 2 ===
Table 3 shows the variety of invoice currencies across exporter continents. In all
continents, local currency plays a minor role as an invoice currency. When importing
from Africa, the share of vehicle currency, maybe either USD or euro (EUR), is
significantly high. The high share of producer currency in importing from America will
be due to the frequent use of USD in importing from the U.S. The similar applies to the
high share of vehicle currency in Asia. The significant share of producer and vehicle
currencies in Europe will be due to the use of EUR and USD, respectively. The
observations in Pacific are interesting. Not only vehicle currency but also producer
currency have the high share in terms of frequency. Namely, Australian dollar (AUD)
and New Zealand dollar (NZD) play some role in invoicing in Thai import from Pacific.
=== Table 3 ===
Next, Table 4 shows the basic statistics on the import transaction-level frequency
and average values according to invoice currencies. As shown in Figure 2, regardless of
invoice currencies, the median of import frequency is one, implying that most of Thai
15
importers trade once a year. In maximum cases, importers trade every day and the
frequency reaches even to 365 times in a year. The mean of frequency in imports
invoiced in local currency is significantly highest among three types of invoicing. On
the other hand, the mean of imports per shipment shows the reverse order. While the
mean in vehicle currency is largest, the case of local currency has the smallest mean of
imports per shipment. The same is true for the median of imports per shipment. Also,
the standard deviation of values per shipment in addition to their maximum is
significantly large in the case of vehicle currency.
=== Table 4 ===
The means of import frequency and imports per shipment are reported by invoice
currency and industry in Table 5. There are four noteworthy points. First, the means of
import frequency and imports per shipment are significantly different across industries.
For example, live animals, vegetable products, and transport equipment have relatively
high frequency. Second, in almost all industries, the mean of import frequency is highest
in local currency invoicing. The exception includes paper products, textiles, footwear,
plastic or glass products, miscellaneous, and art products. Third, in contrast, the mean of
imports per shipment is smallest in local currency invoicing. The exception in the case
of import frequency can be also applied to the case of imports per shipment. In those
industries, local currency invoicing has the largest imports per shipment.
=== Table 5 ===
Last, Table 6 reports the means of import frequency and imports per shipment by
exporter continent. In this table, the findings on import frequency are not so uniform
compared with the previous table for industries. Namely, its mean in local currency
invoicing is highest only in Asia and Europe. The frequency of import from the other
continents is not necessarily highest in local currency invoicing. This tendency will be
consistent with the findings in Table 3. Also, it might be worth noting that the mean
frequency of import from Asian countries is relatively higher in any invoice currencies.
This fact will indicate the significant importance of gravity factors (i.e., geographical
distance between trading partners) for frequency determination, as shown in Hornok and
Koren (2014). On the other hand, the mean of imports per shipment is consistently
largest in vehicle currency invoicing.
16
=== Table 6 ===
4. Empirical Analysis Using the customs data in Thailand, this section empirically investigates
Propositions 1 and 2 specified in Section 2. We first examine how invoice currencies are
related to the frequency of shipments. Specifically, we estimate the following simple
equation.
ln 22
represents the frequency of firm f’s import of HS eight-digit
commodity p from country i in year t.5 is our main variable, which takes the
value one if invoice currency in the concerned transactions is THB and zero otherwise.
is a vector of time-variant import firm characteristics. is fixed effects for the
combination of export country, HS eight-digit commodity, and year. As shown in the
previous section, the mean of values per shipment differs by industries and exporter
continents. The above fixed effects will contribute to controlling for these differences.
We estimate this equation by ordinary least square (OLS) method.
As time-variant import firm characteristics, we introduce two variables. One is
firm f’s total imports from the world in year t. We expect that this variable is related to
the size of import firms. This variable is constructed by aggregating the customs data of
all imports by firms and years. The other is a dummy variable that takes the value one if
firm f gets engaged in exporting activities and zero otherwise. As demonstrated in the
previous studies on frequency, firms also optimize the frequency of export. Therefore,
exporters may have significantly different frequency of import. This variable is
constructed by employing the customs data on export. Those can be integrated by using
firm identification code as a key.6
There are two empirical issues. First, some firms import the same commodity
from the same country under multiple invoice currencies within a year. This observation
may be due to their import from different export firms or due to the change of invoice
currency during the year. Since we cannot identify either one, such observations, which
account for around 10% of all observations, are dropped from estimation sample.
Second, as demonstrated in Section 2, the invoice currency and the frequency are
simultaneously determined. Therefore, the error term is likely to be correlated with local
5 More exactly, as in the previous section, we also identify using the information on quantity unit, of which a subscript is omitted in this section. 6 The basic statistics for our variables are provided in Appendix B.
17
currency dummy variable. To correct this endogeneity bias, we also estimate the above
equation by instrument variable (IV) method. As an instrument, we use the share of
exports under local currency invoicing in total exports. In the case of non-exporters, this
share is set to zero. As demonstrated theoretically and empirically in Chung (2014),
invoice currency in import is highly correlated with that in export. However, the share
of exports under local currency will be not directly related with the frequency of import.
This variable is also constructed by employing the customs data on export.
Table 7 presents the baseline result on the relation between invoice currencies and
import frequency represented by the model (22). Columns (I) and (II) report the results
by the OLS. The models without and with a vector of time-variant import firm
characteristics are respectively estimated. The coefficient for Local Currency Dummy is
estimated to be significantly in both models. Specifically, the frequency of import
shipments is 9-14% higher when those shipments are invoiced in the local currency, i.e.
THB. As implied in Proposition 1, this result indicates that importers can make frequent
orders without consideration on the risk of exchange-rate fluctuations when transactions
are invoiced in their home currency. While the coefficient for Total Imports is
significantly positive, Exporter Dummy has a significantly negative coefficient. These
results indicate that the smaller-sized importers in terms of import values or the
exporters have the lower frequency of import transactions.
=== Table 7 ===
As mentioned above, the invoice currency and frequency of import transactions
are simultaneously determined. To correct the endogeneity biases from this simultaneity,
we estimate the model by the IV method and report the results in columns (III) and (IV).
The Cragg-Donald Wald F and Kleibergen-Paap Wald rk F test statistics, which are used
for the test of weak instruments as a null hypothesis, have significantly high values.
Also, in the first stage regression, the coefficients for THB Export Share are estimated
to be significantly positive, indicating that the importers who intensively use THB in
their exporting are likely to use THB also in importing, as is consistent with the results
in Chung (2014). The results in the second stage regression are qualitatively unchanged
with those reported in columns (I) and (II). The result in column (IV) shows the import
frequency is 26% higher when it is invoiced in the local currency.
We conduct some more robustness checks on the above results. First, we restrict
to import transactions that exist in the previous year. Since we count the number of
dates with import transactions from January to December in each year, we may
18
underestimate the frequency in firms who for the first time start to import on, say,
November. To avoid this underestimation, the estimation sample is restricted only to
import firm-export country-commodity observations that exist in the previous year. The
result is reported in column (I) in Table 8. Second, as explained in Section 3, we have
used import transaction data aggregated at a daily basis (called daily data). We relax this
restriction and define the raw number of import transactions as import frequency (called
full data). The result is provided in column (II). Both columns (I) and (II) show
qualitatively unchanged results with Table 7. Third, we differentiate non-local currency
between vehicle currency and producer currencies because the import frequency may be
systematically different between two kinds of currency. To do that, we introduce two
dummy variables on the use of those currencies as invoice currencies (Vehicle Currency
Dummy and Producer Currency Dummy) instead of Local Currency Dummy (and use
the daily data). The results by the OLS method are shown in columns (III) and (IV) and
show the significantly higher import frequency when the invoice currency is home
currency.
=== Table 8 ===
Next, we empirically investigate how invoice currencies are related to the average
values per shipment, i.e., Proposition 2. To do that, we again estimate the following
simple equation.
ln 23
represents the annual average values per shipment in firm f’s import of HS
eight-digit in year t, which is computed by dividing total annual imports by annual
import frequency. Due to our use of import values in dependent variable, import firm
characteristics do not include firms’ total imports. We use the daily data. The estimation
results by the OLS and the IV are reported in columns (I) and (II) in Table 9,
respectively. The coefficients for Local Currency Dummy are estimated to be
significantly negative, as implied in Proposition 2. Namely, the value per shipment is
significantly smaller for import transactions invoiced in the home currency. However,
the absolute magnitude of its coefficient looks too large in column (II), showing the
471% smaller average shipment of import transactions invoiced in local currency.
=== Table 9 ===
The other results are as follows. First, the coefficients for Exporter Dummy are
19
estimated to be significantly positive, indicating that the exporters have the larger
average shipment values of import transactions. Second, the estimation result of the
equation with Vehicle Currency Dummy and Producer Currency Dummy instead of
Local Currency Dummy is reported in column (III). Consistent with the result for Local
Currency Dummy, both those two dummy variables have the significantly positive
coefficients. Jointly with the above findings on the frequency, these contrasting results
between the frequency of import transactions and their average shipment values imply
that importers make just-in-time orders and shipments when the import transactions are
invoiced in their own currency. Our model provided a rationale for these results by
shedding the light on the role of the cost of exchange-rate risk management.
5. Benefits of an Invoice Currency In this section, using equations (17)-(21), we compute per-shipment fixed costs
for FCI management ( ). To do that, we need to specify three parameters, i.e., , , ,
and . The elasticity of substitution between varieties ( ) is obtained from Broda and
Weinstein (2006), which provides the elasticity at an HS three-digit level for Thailand.
We use the same values for gross ad valorem storage costs ( ) and discount rate ( ) as
Kropf and Sauré (2014), which are respectively 0.35 and 0.05. The forward premium
offered by a financial institution to each firm is an important parameter in our analysis
but unobservable. We simply assume that this parameter can be well approximated by
the change rate of exchange rates with THB from the current to the next year (average).
Therefore, this parameter is defined at an export country-year level.
Applying these parameters, we can compute , i.e., or , for each
import transaction. Importantly, we cannot observe both cases of HCI and FCI for each
transaction. Namely, we infer from import transactions under HCI and from
those under FCI. To obtain , we estimate the following simple equation.
24
We re-label subscripts for ; , , , and stand for import firm, export country, HS
eight-digit code, and year, respectively. is equal to for the case of HCI and
for the case of FCI. takes the value one if invoice currency in the
concerned transaction is foreign currency (i.e., not THB) and zero otherwise. In this
estimation, coefficient shows the fixed costs for FCI management ( ). We also
include year and export country fixed effects.
The estimation results are reported in Table 10. For rescale, we divide by one
thousand. Coefficient is shown as “Difference”. In column (I), we do not include
20
any fixed effects. The constant term is estimated to be 27, indicating that the average
per-import shipment fixed costs are 27 thousand THB (approximately 800 USD). This
magnitude is one-tenth of the average per-export shipment fixed costs in Swiss
(5,723CHF). Therefore, we may say that fixed costs per-import shipment are much
lower than those per-export shipment. The “difference” is 51, which indicates that the
average per-shipment fixed costs for FCI management are 51 thousand THB
(approximately 1,500 USD) and are around twice higher than the average per-import
shipment fixed costs. This magnitude of the difference is unchanged even if controlling
for year fixed effects, as shown in column (II). However, when controlling for export
country fixed effects, it becomes 119 thousand THB (approximately 3,600 USD). This
change will be because export country fixed effects control for the differences in
per-shipment import fixed costs ( ) across export countries and such differences are
significant.
=== Table 10 ===
Next, we examine the correlation of fixed costs for FCI management with export
country characteristics. To do that, by estimating equation (24) by export country (no
fixed effects), we infer those costs by export country, of which estimates are available in
Appendix. Then, we regress a log of those costs on various characteristics of export
country in year 2007. Those include gravity variables, namely, GDP, GDP per capita,
and the geographical distance with Thailand. Also, we introduce RTA dummy variable,
which takes the value one for partner countries of Thailand’s RTAs.7 Our main variable
in this estimation is Turnover Share, which is the share of the export country currency in
daily global foreign exchange market turnover, of which data are obtained from the BIS
Triennial Central Bank Survey in 2007. Currencies not listed in the survey are assigned
zero shares. As suggested in Goldberg and Tille (2011) and Chung (2014), its higher
share values will indicate the lower transaction costs for the export country currency.
The estimation results are reported in Table 11. Based on the high correlations
between GDP and GDP per capita and between Distance and RTA Dummy, we also
estimate the equations separately including those variables. There are three noteworthy
points. First, the export country’s GDP, GDP per capita, and distance with Thailand do
not affect the fixed costs for FCI management. Second, those costs are lower in
importing from RTA partners. The majority of the RTA partners are ASEAN plus three
7 As of 2007, RTA partners include ASEAN countries, Australia, China, India, Japan, and New Zealand. We also include Korea, with which the RTA was entry into force in 2010.
21
countries (i.e., China, Japan, and Korea). Among those countries, there exist some
currency cooperation schemes such as Chiang Mai Initiative. Therefore, the significant
correlation may come from the effects of such schemes rather than the effects of RTAs.
Last, as is consistent with our expectation, the higher turnover share of export country
currency is associated with the lower fixed costs for FCI management.
=== Table 11 ===
6. Concluding Remarks In this study, we examined the benefits of home currency invoicing, which are
represented by fixed costs of exchange rate risk management paid by importers when
they work with foreign currency invoicing. We revealed that those benefits for average
Thai importer are significantly positive, and range between 7.3% (1,500 USD) of
one-time shipment value and 17.1% (3,600 USD). Further, those benefits become larger
when Turnover Share is lower, or the export country is one of Thailand’s RTA partners.
It is also found that frequency of shipments is higher and the value per shipment is
smaller for import transactions priced in buyers’ currency than those not priced in it.
Our results propose clear qualitative policy implication on internationalization of
currencies from a microeconomic point of view.
22
Appendix A. Constancy of the Price Index
The price index is rewritten as
, A1
where
≡ , A2
≡ . A3
According to (7), prices of final goods which are produced at and sold at are
written as
1
∗
. A4
Thus, is rewritten as
1
1∗ . A5
For HCI importers, we proved that the equilibrium interval does not depend on , which
implies
∆ . A6
Thus,
1
1∗ ∆ 1 , A7
and does not depend on and constant over time.
Similarly, is rewritten as follows:
1
1∗ . A8
Arranging producer indices so that ∆ is increasing in . For 0, consider the
distribution where ∈ , is shipped at time . For given , the date of the
last shipment is derived as
∆ max ∈ | ∆ . A9
Here, is the set of integers. Suppose importers which make shipments at time . The
23
following relation holds for these importers:
∆ . A10
Define the solution by , , . The solution is uniquely determined as ∆ is
increasing in . Thus, the following relation holds for ∈ , , , 1, , ∩
Ω , where Ω is the set of FCI importers:
∆ , , ∆ , , . A11
Further,
∆ , , ∆ , , . A12
(A10) leads to lim → , , lim → 1, , , which jointly implies with
(A12) that does not depend on in this limit. According to (A8),
does not depend on and constant over time, and is so, too.
Appendix B. Other Tables
Table B1. Basic Statistics
Obs Mean Std. Dev. Min Max
ln Frequency (Daily) 4,980,162 0.5900 0.9095 0 5.8972
ln Frequency (Full) 4,980,162 0.8156 1.1362 0 12.0622ln Values 4,980,162 10.2618 2.4298 0 22.8140Local Currency Dummy 4,980,162 0.0601 0.2377 0 1ln Total Imports 4,980,162 17.6609 3.1781 2.5649 26.4119Exporter Dummy 4,980,162 0.6724 0.4693 0 1THB Export Share 4,980,162 0.0981 0.2571 0 1Vehicle Currency Dummy 4,980,162 0.5304 0.4991 0 1Producer Currency Dummy 4,980,162 0.4095 0.4917 0 1
Source: Authors’ computation
24
Table B2. Estimates on Fixed Costs by Export Country
Obs.Coef. S.D. Coef. S.D.
AFG 25 24 2 22 395
AGO 6,697 14,472 13 13,939 180ALB 6 13 9 13 159ARE 515 156 4 129 10,705ARG 533 307 14 300 3,646ARM 104 101 4 90 56ATG 70 242 1 240 70AUS 120 38 10 35 83,867AZE 14,496 19,180 3 17,824 88BDI 283 249 3 213 15BEN 585 769 13 765 95BGD 171 129 3 122 2,721BGR 86 102 5 98 2,328BHR 1,722 554 3 413 1,365BHS 77 79 3 78 76BIH 6 12 5 12 150BLR 1,958 1,956 1 1,862 160BLZ 841 2,147 67 2,033 251BMU -1 2 3 2 43BOL 54 63 15 61 85BRA 648 239 32 226 11,068BRB 165 680 0.2 674 56BRN 2,922 2,268 3 2,066 778BTN 2,374 1,389 2 1,091 141BWA 798 512 1 481 69CAN 114 64 15 62 29,634CHE 49 13 12 13 76,782CHL 175 186 42 183 1,810CHN 14 8 28 8 937,035CMR 358 773 8 757 359COG 385 211 4 176 146COK 207 597 6 586 28COL 317 722 4 700 1,000COM 38 15 3 9 13CRI 47 48 5 47 1,186CZE 41 12 9 12 13,914DJI 65 36 6 27 43DMA -9 76 85 73 73DNK 28 8 12 7 29,479
Difference Constant
Note: “Coef.”, “S.D.” and “Obs.” are coefficient estimate, standard deviation, and the number of
observations, respectively.
25
Table B2. Estimates on Fixed Costs by Export Country (Cont.)
Obs.Coef. S.D. Coef. S.D.
DOM 18 20 21 19 688
DZA 27,889 24,927 208 22,641 80EGY 311 200 6 185 1,729ERI 54 55 0.4 50 41ETH 83 120 6 114 320FJI 134 54 2 45 165GAB 7,210 6,126 18 5,255 106GBR 53 11 9 11 155,685GEO 137 103 5 97 457GHA 354 171 3 160 190GIN 109 161 1 157 78GMB 101 97 2 91 80GTM 100 65 2 64 298HKG 35 16 9 16 105,788HND 18 20 14 20 290HRV 126 549 17 539 646HUN 71 63 6 57 8,644IDN 151 60 27 58 59,265IND 88 36 16 35 84,948IRN 1,521 835 10 759 1,231IRQ 2,208 7,532 3 7,395 305ISL 298 173 2 164 609ISR 77 48 16 48 14,906JAM 70 39 35 35 197JOR 181 134 6 123 553JPN 2 4 45 4 886,728KAZ 955 3,053 191 3,037 192KEN 34 13 3 12 1,033KGZ 5,395 7,659 14 6,434 34KHM -5 6 33 5 3,513KNA 36 133 9 128 14KOR 79 23 27 23 177,722KWT 2,427 774 3 627 1,114LAO -16 8 73 6 7,126LBN 101 62 9 55 312LBR 81 63 1 55 46LBY 38,538 35,732 3 30,565 41LKA 38 37 4 36 4,116LSO 37 108 14 105 34
Difference Constant
Note: “Coef.”, “S.D.” and “Obs.” are coefficient estimate, standard deviation, and the number of
observations, respectively.
26
Table B2. Estimates on Fixed Costs by Export Country (Cont.)
Obs.Coef. S.D. Coef. S.D.
LTU 54 105 2 102 921
LVA 44 19 2 19 529MAC 11 21 10 20 759MAR 197 446 7 434 2,568MDA 4 6 3 6 223MDG 16 14 2 14 601MDV 345 98 2 86 331MEX 81 92 18 90 18,819MKD 33 100 1 99 174MLI 163 128 4 126 310MMR 3,374 1,831 15 1,075 7,085MNG 1,048 2,638 24 2,509 73MNP 182 258 5 214 19MOZ 249 125 1 122 294MRT 88 87 62 83 82MSR 15 39 7 38 16MUS 29 24 5 23 709MWI 320 246 3 237 76MYS 63 26 30 25 140,239NAM 188 183 9 176 236NCL 147 52 3 39 84NER 82 28 5 20 47NGA 1,559 1,528 4 1,433 898NIC 4 90 23 89 204NIU -2 2 3 2 45NOR 124 107 78 98 7,124NPL 10 10 3 10 1,347NRU 314 480 5 440 25NZL 81 35 6 33 13,263OMN 3,668 803 3 566 1,080PAK 110 106 10 101 6,018PAN 1,423 2,832 164 2,767 267PER 276 593 6 581 1,173PHL 139 101 16 97 26,986PNG 1,912 2,165 6 2,046 394POL 61 29 12 27 9,064PRI -246 71 349 67 767PYF -209 71 271 65 172QAT 3,374 934 23 676 1,689
Difference Constant
Note: “Coef.”, “S.D.” and “Obs.” are coefficient estimate, standard deviation, and the number of
observations, respectively.
27
Table B2. Estimates on Fixed Costs by Export Country (Cont.)
Obs.Coef. S.D. Coef. S.D.
ROM 52 49 9 47 3,821
RUS 2,649 1,045 4 1,001 2,996SAU 2,287 635 6 488 3,970SDN 1,814 3,858 2 3,774 443SEN 297 200 8 190 140SGP 46 12 17 11 198,006SLB 1,473 2,249 7 2,240 123SLE 42 38 1 36 97SLV 276 1,747 14 1,745 372STP 115 446 1 440 35SUR 121 177 0.3 174 30SWE 70 16 16 15 41,393SWZ 108 112 67 106 1,820SYC 1,295 962 73 872 163SYR -5 47 72 41 137TCA 98 286 1 282 36TCD 1,807 4,516 14 4,370 47TGO 393 324 1 317 90TJK 344 244 22 207 25TKL 51 61 8 60 158TON 8 6 3 4 11TTO 148 126 1 121 81TUN 20 8 2 7 2,084TUR 165 231 9 224 14,888TUV 1,472 1,214 0 1,184 41TWN 3 7 40 7 311,303TZA 104 53 5 51 1,023UGA 155 75 3 70 150UKR 3,997 3,163 9 3,062 1,356URY -358 203 539 201 364USA 40 9 13 9 511,402VEN 881 893 4 760 213VGB 88 174 48 170 80VNM 101 61 12 59 35,566WLF 17 21 0.1 20 13WSM 64 181 7 178 34YEM 9,170 7,997 8 7,706 266ZAF 260 91 6 86 7,924ZMB 657 1,030 4 1,019 198
Difference Constant
Note: “Coef.”, “S.D.” and “Obs.” are coefficient estimate, standard deviation, and the number of
observations, respectively.
28
References
Alessandria, G., Kaboski, J.P., and Midrigan, V., 2010, Inventories, Lumpy Trade, and
Large Devaluations, American Economic Review, 100 (5), 2304-2339.
Anderson. J.E. and van Wincoop, E., 2004, Trade Costs, Journal of Economic Literature,
42, 691–751.
Bekes, G., Fontagne, L., Murakozy, B., and Vicard, V., 2014, Shipment Frequency of
Exporters and Demand Uncertainty, CESifo Working Paper No. 4734.
Cherkashin, I., Demidova, S., Kee, H., and Krishna, K., 2015, Firm Heterogeneity and
Costly Trade: A New Estimation Strategy and Policy Experiments, Forthcoming
in Journal of International Economics.
Chung, W., 2014, Imported Inputs and Invoicing Currency Choice: Theory and
Evidence from UK Transaction Data, Discussion Papers 2014-11, University of
Nottingham, GEP.
Das, S., Roberts, M., and Tybout, J., 2007, Market Entry Costs, Producer Heterogeneity,
and Export Dynamics, Econometrica, 75(3), 837-873.
Eaton, J., Eslava, M., Kugler, M., and Tybout, J., 2008, Export Dynamics in Colombia:
Transaction-level Evidence, In: E. Helpman, D. Marin, and T. Verdier (Eds.), The
Organization of Firms in a Global Economy, Cambridge, MA: Harvard
University Press, pp. 231-272.
Fabling, R., and Sanderson, L., 2015, Export Performance, Invoice Currency and
Heterogeneous Exchange Rate Pass-through, The World Economy, 38 (2),
315-339.
Gopinath, G., Itskhoki, O., and Rigobon, R., 2010, Currency Choice and Exchange Rate
Pass-Through, American Economic Review, 100 (1), 304-336.
Goldberg, L. S., and Tille, C., 2008, Vehicle Currency Use in International Trade,
Journal of International Economics, 76 (2), 177-192.
Goldberg, L. S., and Tille, C., 2009, Micro, Macro, and Strategic Forces in International
Trade Invoicing, NBER Working Paper No.15470.
Hornok, C. and Koren, M., 2015, Per-shipment Costs and the Lumpiness of
International Trade, Forthcoming in Review of Economics and Statistics.
Kropf, A. and Sauré, P., 2014, Fixed Costs per Shipment, Journal of International
Economics, 92(1), 166-184.
Reiss, D. G., 2015, Invoice Currency: Puzzling Evidence and New Questions from
Brazil, Banco Central do Brasil Working Papers No.382.
Rose, A. and van Wincoop, E., 2001, National Money as a Barrier to International
29
Trade: The Real Case for Currency Union, American Economic Review, 91(2),
386-390.
30
Table 1. The Decomposition of Import Transactions According to Invoice Currencies
Number/Value Share Number/Value Share Number/Value Share
Import Transactions
2007 315,220 0.07 1,810,002 0.39 2,518,534 0.542008 345,968 0.07 1,910,047 0.38 2,788,492 0.552009 354,500 0.07 1,678,244 0.35 2,762,252 0.582010 427,163 0.08 1,967,099 0.35 3,288,345 0.582011 479,208 0.08 2,017,907 0.34 3,406,764 0.58
Import Values (Million THB)2007 210,254 0.04 987,052 0.20 3,645,962 0.752008 220,038 0.04 1,082,668 0.19 4,467,971 0.772009 182,519 0.04 825,785 0.19 3,385,206 0.772010 255,404 0.05 1,042,832 0.18 4,340,484 0.772011 289,262 0.04 1,196,530 0.18 5,129,073 0.78
Local Producer Vehicle
Source: Authors’ computation
Table 2. Import Frequency and Imports per Shipment by Invoice Currency and Industry
Frequency Value Frequency Value Frequency Value
Live animals 0.08 0.02 0.45 0.19 0.47 0.79
Vegetable products 0.07 0.02 0.24 0.22 0.69 0.76Animal/vegetable fats and oils 0.03 0.02 0.45 0.08 0.52 0.90Food products 0.05 0.10 0.47 0.20 0.49 0.71Mineral products 0.08 0.00 0.40 0.01 0.51 0.98Chemical products 0.10 0.10 0.37 0.18 0.53 0.72Plastics and rubber 0.07 0.05 0.38 0.30 0.55 0.65Leather products 0.07 0.06 0.18 0.21 0.75 0.73Wood products 0.14 0.05 0.34 0.18 0.52 0.77Paper products 0.08 0.03 0.39 0.21 0.53 0.76Textiles 0.07 0.02 0.19 0.20 0.74 0.77Footwear 0.04 0.02 0.17 0.05 0.79 0.93Plastic or glass products 0.08 0.04 0.37 0.35 0.54 0.61Precision metals 0.03 0.00 0.30 0.11 0.66 0.88Base Metal 0.09 0.03 0.41 0.18 0.50 0.80Machinery 0.08 0.06 0.31 0.25 0.61 0.69Transport equipment 0.11 0.12 0.40 0.34 0.49 0.54Precision machinery 0.08 0.09 0.43 0.41 0.49 0.50Miscellaneous 0.07 0.05 0.24 0.30 0.68 0.66Art products 0.26 0.12 0.31 0.47 0.43 0.42
Local Producer Vehicle
Source: Authors’ computation
31
Table 3. Import Frequency and Imports per Shipment by Invoice Currency and Exporter
Continent
Frequency Value Frequency Value Frequency Value
Africa 0.10 0.01 0.01 0.00 0.89 0.99
America 0.06 0.04 0.77 0.65 0.17 0.31Asia 0.07 0.04 0.25 0.12 0.69 0.84Europe 0.14 0.08 0.49 0.29 0.37 0.64Pacific 0.09 0.01 0.50 0.06 0.41 0.93
Local Producer Vehicle
Source: Authors’ computation
Table 4. Basic Statistics for Import Frequency and Imports per Shipment by Invoice
Currency
Mean S.D. Median Maximum
Frequency
Local 5.559 16.224 1 331Producer 4.227 11.324 1 365Vehicle 5.174 14.359 1 364Total 4.831 13.416 1 365
Values per shipment (Thousand THB)Local 282 2,494 10 254,875Producer 421 6,169 27 1,961,622Vehicle 947 25,995 37 7,805,302Total 695 19,479 30 7,805,302
Source: Authors’ computation
32
Table 5. The Means of Import Frequency and Imports per Shipment by Invoice
Currency and Industry
Local Producer Vehicle Local Producer Vehicle
Live animals 10.718 9.728 5.289 301 1,515 3,124
Vegetable products 11.613 5.922 8.768 493 1,421 1,658Animal/vegetable fats and oils 4.878 3.913 4.439 436 262 2,681Food products 7.496 4.045 4.416 1,099 354 1,466Mineral products 7.718 3.987 4.663 1,117 762 45,256Chemical products 9.858 4.577 4.761 1,065 439 1,407Plastics and rubber 9.533 5.761 5.746 192 171 306Leather products 7.406 2.624 5.017 111 98 226Wood products 6.634 4.926 3.516 129 178 524Paper products 2.764 3.091 3.926 39 90 274Textiles 2.110 2.917 5.145 45 150 304Footwear 4.213 2.240 5.315 46 39 123Plastic or glass products 2.174 3.845 4.085 37 200 248Precision metals 7.933 6.181 5.315 453 1,640 4,886Base Metal 8.488 4.914 4.784 169 246 971Machinery 5.871 3.952 5.709 380 621 509Transport equipment 10.516 6.104 6.319 672 829 1,707Precision machinery 6.153 3.455 3.903 467 344 315Miscellaneous 1.776 2.495 3.806 40 126 153Art products 1.013 1.155 1.302 12 119 85
Frequency Values per shipment (Thousand THB)
Source: Authors’ computation
Table 6. The Means of Import Frequency and Imports per Shipment by Invoice
Currency and Exporter Continent Local Producer Vehicle
Frequency
Africa 2.835 1.309 4.218America 3.102 4.001 3.569Asia 6.267 5.304 5.576Europe 6.110 3.219 3.805Pacific 1.313 2.924 3.413
Values per shipment (Thousand THB)Africa 101 38 4,703America 183 492 1,599Asia 364 426 884Europe 215 380 688Pacific 75 311 4,009
Source: Authors’ computation
33
Table 7. Determinants of Import Frequency
(I) (II) (III) (IV)
Second stage
Local Currency Dummy 0.0893*** 0.1328*** 0.6000*** 0.2310***[0.0023] [0.0022] [0.0133] [0.0124]
ln Total Imports 0.0763*** 0.0766***[0.0002] [0.0002]
Exporter Dummy -0.0618*** -0.0620***[0.0011] [0.0010]
Number of Observations 4,980,162 4,980,162 4,744,619 4,744,619R-squared (Centered) 0.1948 0.2354 -0.0131 0.050
First stageTHB Export Share 0.1293*** 0.1418***
[0.0006] [0.0006]Centered R-squared 0.027 0.0321Cragg-Donald Wald F 1.20E+05 1.40E+05Kleibergen-Paap Wald rk F 44547.07 5.10E+04
IVOLS
Notes: The dependent variable is a log of the frequency of import shipments. The parentheses are
robust standard errors. ***, **, and * indicate 1%, 5%, and 10% significance, respectively. We
include fixed effects for the combination of export country, HS eight-digit commodity, and year. In
the results for the first stage regression of IV method, we report the result of the excluded variable,
THB Export Share, which is the share of exports under local currency invoicing in total exports.
Various test statistics are also presented.
34
Table 8. Robustness Checks on Import Frequency
(I) (II) (III) (IV)Estimation Method IV IV OLS OLSData type Daily Full Daily Daily
Second stage
Local Currency Dummy 0.2870*** 0.2804***[0.0202] [0.0156]
Vehicle Currency Dummy -0.0743*** -0.1511***[0.0024] [0.0024]
Producer Currency Dummy -0.1006*** -0.1193***[0.0023] [0.0023]
ln Total Imports 0.0930*** 0.0905*** 0.0766***[0.0004] [0.0002] [0.0002]
Exporter Dummy -0.0633*** -0.0616*** -0.0620***[0.0021] [0.0013] [0.0011]
Number of Observations 1,632,779 4,744,619 4,980,162 4,980,162R-squared (Centered) 0.0444 0.045 0.1948 0.2355
First stageTHB Export Share 0.1657*** 0.1418***
[0.0011] [0.0006]Centered R-squared 0.0283 0.0321Cragg-Donald Wald F 9.08E+04 1.40E+05Kleibergen-Paap Wald rk F 24713.77 51233.44
Notes: The dependent variable is a log of the frequency of import shipments. The parentheses are
robust standard errors. ***, **, and * indicate 1%, 5%, and 10% significance, respectively. We
include fixed effects for the combination of export country, HS eight-digit commodity, and year. In
the results for the first stage regression of IV method, we report the result of the excluded variable,
THB Export Share, which is the share of exports under local currency invoicing in total exports.
Various test statistics are also presented.
35
Table 9. Determinants of Average Import Values per Shipment
(I) (II) (III)Estiamtion Method OLS IV OLS
Second stage
Local Currency Dummy -0.5679*** -1.7425***[0.0050] [0.0291]
Vehicle Currency Dummy 0.6341***[0.0054]
Producer Currency Dummy 0.5179***[0.0052]
Exporter Dummy 0.3248*** 0.3144*** 0.3215***[0.0024] [0.0023] [0.0024]
Number of Observations 4,980,162 4,806,105 4,980,162R-squared (Centered) 0.3948 -0.0054 0.3950
First stageTHB Export Share 0.1431***
[0.0006]Centered R-squared 0.0312Cragg-Donald Wald F 1.40E+05Kleibergen-Paap Wald rk F 52865.43
Notes: The dependent variable is a log of the average import values per shipment. The parentheses
are robust standard errors. ***, **, and * indicate 1%, 5%, and 10% significance, respectively. We
include fixed effects for the combination of export country, HS eight-digit commodity, and year. In
the results for the first stage regression of IV method, we report the result of the excluded variable,
THB Export Share, which is the share of exports under local currency invoicing in total exports.
Various test statistics are also presented.
36
Table 10. Fixed Costs for FCI Management
(I) (II) (III)
Difference 51.0769*** 50.8037*** 119.2677***
[7.1360] [7.1412] [8.0531]Constant 26.5147***
[6.8806]Year Dummy NO YES YESExport Country Dummy NO NO YESNumber of Observations 4,171,649 4,171,649 4,171,649
Notes: The parentheses are standard errors. *** indicate 1% significance.
Table 11. Correlation of Fixed Costs for FCI Management with Export Country
Characteristics
(I) (II) (III) (IV) (V)
ln GDP 0.0524 0.0618 0.0549 0.0102
[0.0915] [0.0793] [0.0915] [0.0865]ln GDP per capita 0.027 0.0622 0.0213 0.0256
[0.1184] [0.1015] [0.1145] [0.1147]ln Distance -0.1797 -0.1876 -0.1749 0.0722
[0.3196] [0.3159] [0.3115] [0.2483]RTA Dummy -1.2538* -1.1481* -1.2528* -1.0567*
[0.6827] [0.6182] [0.6825] [0.5540]Turnover Share -6.3278** -5.7026** -6.2606** -6.6050** -6.2948**
[2.7204] [2.4059] [2.6892] [2.7919] [3.1262]Constant 5.1765 6.2658** 5.1945 3.5555* 3.9992
[3.4003] [2.7313] [3.4133] [2.1429] [3.2014]Number of Observations 137 137 137 137 137R-squared 0.0402 0.0376 0.0398 0.0375 0.0159
Notes: The parentheses are robust standard errors. ***, **, and * indicate 1%, 5%, and 10%
significance, respectively.
37
Figure 1. Structure of the Model (Simplified Three-Country Example)
Figure 2. Import Frequency and Invoice Currency
38
Figure 3. Sample Distribution of Number of Shipments per Year
Source: Authors’ compilation
02
46
De
nsity
0 2 4 6Number of Shipments per Year, logged
39
Figure 4. Sample Distribution of Average Import Values per Shipment
Source: Authors’ compilation
0.0
5.1
.15
.2D
ens
ity
0 5 10 15 20 25Average Values in THB, logged